Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
144 lines (108 loc) · 4.47 KB

CONTRIBUTING.md

File metadata and controls

144 lines (108 loc) · 4.47 KB

mulmocast-cli

Initialization

Install dependencies.

yarn install
brew install ffmpeg # Or install ffmpeg from https://ffmpeg.org/download.html

Create .env file.

OPENAI_API_KEY=your_openai_api_key
GOOGLE_PROJECT_ID=your_google_project_id # optional for movie, image generation
NIJIVOICE_API_KEY=your_nijivoice_api_key # optional for movie, audio generation
BROWSERLESS_API_TOKEN=your_browserless_api_token # optional for scripting from web content

Quick Start

Generate MulmoScript interactively

yarn run scripting -i -t seed_interactive

This will create a MulmoScript JSON file in the ⁠./output folder by default.

Generate content using either approach:

  • Option 1: Execute each step individually

    yarn run audio scripts/test/test_en.json
yarn run images scripts/test/test_en.json
yarn run movie scripts/test/test_en.json

  • Option 2: Generate video in one go (this will process audio and images internally)

    yarn run movie scripts/test/test_en.json

Create a podcast episode

  1. Create a MulmoScript (LLMs are useful for this process)
    1. Feed some source text (ideas, news, press releases) to your favorite LLM.
    2. Ask the LLM to write a podcast script in MulmoScript JSON format (you can use the prompt examples in the prompts folder as a reference).
    3. Save the generated script as a MulmoScript JSON file (such as graphai_intro.json)
  2. Run yarn run audio {path to the script file}.
  3. The output will be generated in the ./output folder.

Create a video

  1. Create a script (MulmoScript) using Claude or another LLM
  2. If you need a vertical video for YouTube Live, add those instructions manually
  3. Generate audio files using yarn run audio {path to the script file}
  4. Generate image files using yarn run images {path to the script file}
  5. Generate the final video file using yarn run movie {path to the script file}

MulmoScript format

{
  "title": "title of the podcast",
  "description": "The description of the podcast.",
  "reference": "URL to the source data", // optional
  "speechParams": {
    "provider": "openAI", // or "nijivoice", default is "openAI"
    "speakers": {
      "Host": {
        "voiceId": "shimmer",
        "displayName": {
          "en": "Host"
        }
      },
      "Guest": {
        "voiceId": "echo",
        "displayName": {
          "en": "Guest"
        }
      }
    }
  },
  "beats": [
    {
      "speaker": "Host",
      "text": "words from the host."
    },
    {
      "speaker": "Guest",
      "text": "words from the guest."
    },
    ...
  ]
}

Translate the script into Japanese

Run yarn run translate {path to the script file}

Create a movie file with the Japanese script

Run yarn run movie {path to the script file}

Development Notes

Parts of the Mulmocast codebase utilize GraphAI technology. For more details about GraphAI, please visit the GraphAI repository.

For example, in src/actions/images.ts, we implement GraphAI's GraphData to define the data flow for image generation:

const graph_data: GraphData = {
  // GraphAI data flow definition
}

The ⁠GraphAILogger used in this project is also a utility provided by the GraphAI framework, as mentioned in the Logging Guidelines section below.

Logging Guidelines for Developers

Do not use console.xxx functions directly.
Instead, use GraphAILogger to manage user-facing logs and control output based on verbosity flags.

Log Levels

  • Use GraphAILogger.info for messages intended to be shown to users.
  • Use GraphAILogger.log and GraphAILogger.error for developer-level messages — these will only appear when the -v (--verbose) flag is enabled.

Why Use GraphAILogger

  • Allows user-facing logs to be toggled on/off cleanly.
  • Prevents internal logs from interfering with command output or script arguments.
  • Keeps behavior consistent across different CLI tools and environments.

Please ensure you choose the appropriate GraphAILogger method based on the log's purpose and expected visibility.

Pull Requests

All pull requests must pass lint checks, build, and CI tests before they can be merged.

Release

In addition to passing the GitHub CI tests, you must run test/release_test.sh and ensure that all contents are successfully generated.